Comet
PulseAugur coverage of Comet — every cluster mentioning Comet across labs, papers, and developer communities, ranked by signal.
- developed by Perplexity AI 100%
- acquired by Atlas Ai Model 90%
- competes with Safari 80%
- competes with Detroit Institute of Arts 80%
- competes with The Browser Company 70%
- competes with opera 70%
- competes with Brave 70%
- instance of Brave 70%
- competes with Atlas Ai Model 70%
- instance of The Browser Company 70%
- instance of Detroit Institute of Arts 60%
- used by Atlas Ai Model 50%
- 2026-05-24 research_milestone Researchers discovered and disclosed a prompt injection vulnerability in the Perplexity Comet browser. source
8 day(s) with sentiment data
-
LLM observability platforms diverge on advanced features as market booms
The LLM observability and evaluation platform market is rapidly expanding, with projections reaching $9.26 billion by 2030. Platforms are diversifying into AI-native tools, open-source evaluation libraries, AI gateways,…
-
Perplexity AI faces surge of fake apps and malicious extensions in Russia
Perplexity AI is facing a surge of fake applications and malicious extensions targeting users in Russia, particularly during periods of instability with app stores. These fraudulent versions often mimic the official Per…
-
User finds Perplexity's workflow design intuitive
The user found that Perplexity's workflow design aligned well with their own preferences. This suggests a positive user experience with the AI's interface and functionality.
-
US court allows Perplexity AI shopping agents on Amazon, overturning injunction
A US appeals court has ruled in favor of Perplexity, overturning an injunction that previously barred its AI shopping agents from accessing Amazon. The court determined that it is the user, not Perplexity, who accesses …
-
FAU researchers achieve top ranks in ImageCLEF 2026 multimodal reasoning task
Researchers from Florida Atlantic University (FAU) have developed a system for the ImageCLEF 2026 task on Multimodal Reasoning, focusing on Visual Multiple Choice Question Answering (Visual MCQ) and Visual Open Question…
-
Ex-Perplexity employee launches AI browser for knowledge work
A former Perplexity employee, who previously contributed to the Comet project, has launched a new AI-powered browser. This browser is specifically designed to assist with knowledge work, aiming to streamline research an…
-
AI browsers vulnerable to prompt injection attacks, researchers warn
AI browsers, which can act on user accounts and navigate the web, are vulnerable to prompt injection attacks where malicious instructions on a webpage or in an email can trick the AI agent into performing unintended act…
-
Browser wars shift to AI assistants as new alternatives emerge
The browser landscape is evolving beyond traditional search, with a new focus on AI-powered assistants. Companies like Perplexity, The Browser Company, and Opera are launching browsers with integrated AI features design…
-
Macs quietly host free LLMs; Sigma browser leverages local AI
The Sigma browser offers a privacy-focused approach by utilizing a local LLM called Eclipse for its AI chat features, with an option to connect to cloud models for more demanding tasks. This hybrid model aims to balance…
-
OpenAI folds Atlas AI browser into ChatGPT, shifts strategy
OpenAI is discontinuing its Atlas AI browser after less than a year, integrating its features into the updated ChatGPT desktop app and Chrome extension. This strategic shift aims to embed AI browsing capabilities direct…
-
CoPiT pipeline boosts low-resource Mongolian translation accuracy
Researchers have developed CoPiT, a novel translation pipeline designed to address the challenges of low-resource languages, specifically focusing on Mongolian. This system leverages the imbalance in data availability b…
-
New COMET method uses quantum algorithm for gene editing optimization
Researchers have developed a new method called COMET that uses the Quantum Approximate Optimization Algorithm (QAOA) to solve complex combinatorial optimization problems in gene editing. This approach addresses the chal…
-
NAVER LABS re-implements instruction-following pipeline for IWSLT 2026 task
Researchers have re-implemented the NAVER LABS instruction-following pipeline for the IWSLT 2026 Shared Task, adapting it to use SeamlessM4T-v2-large as the speech encoder and Qwen3-4B-Instruct as the LLM backbone. This…
-
New MBR decoding approach incorporates bidirectional effects for improved text generation
Researchers have introduced a novel noisy channel decomposition for Minimum Bayes Risk (MBR) decoding, aiming to improve text generation quality. This approach addresses the asymmetry in common evaluation metrics like B…
-
AI agent integrates with Comet ML for natural language experiment tracking
A developer has created a system using the Model Context Protocol (MCP) to integrate AI agents with experiment tracking platforms like Comet ML. This approach allows ML engineers to query their experiment data using nat…
-
AI-powered browsers challenge Chrome and Safari in new market battle
The browser market is evolving beyond search, with a new focus on AI-powered assistants integrated directly into browsers. While Google Chrome and Apple Safari remain dominant, emerging startups and tech giants are laun…
-
New FLARE-AI website aims to centralize AI harm reporting
A new crowdsourced website called FLARE-AI has been launched by a group of AI researchers to report and track harms caused by artificial intelligence systems. This open-source platform aims to provide a centralized and …
-
New CoMet method improves uncertainty estimation in multimodal LLMs
Researchers have introduced CoMet, a novel method for estimating uncertainty in multimodal large language models (MLLMs). CoMet decomposes uncertainty into context-specific and multiplicity-specific terms, allowing for …
-
New AI models tackle low-resource Tangkhul-English translation
Researchers have developed two neural machine translation systems for the low-resource Tangkhul-English language pair. The primary system, utilizing ByT5-large fine-tuned on over 38,000 parallel sentences, achieved a BL…
-
AI agents challenge browser security models built for humans · 4 sources tracked
The integration of large language models into web browsers presents a significant challenge to existing security models, which are built on the assumption of human users. Features like same-origin policy, user gesture r…